|Publication number||US5086480 A|
|Application number||US 07/594,670|
|Publication date||Feb 4, 1992|
|Filing date||Oct 5, 1990|
|Priority date||May 6, 1987|
|Also published as||CA1308486C, DE3875583D1, EP0313612A1, EP0313612B1, WO1988009101A1|
|Publication number||07594670, 594670, US 5086480 A, US 5086480A, US-A-5086480, US5086480 A, US5086480A|
|Inventors||Graham G. Sexton|
|Original Assignee||British Telecommunications Public Limited Company|
|Export Citation||BiBTeX, EndNote, RefMan|
|Patent Citations (9), Non-Patent Citations (10), Referenced by (76), Classifications (29), Legal Events (3)|
|External Links: USPTO, USPTO Assignment, Espacenet|
This is a continuation of application Ser. No. 295,214, filed Jan. 5, 1989, now abandoned.
This invention relates to analysis and processing of video images.
A video image (which will be understood to encompass frozen images such as facsimile images, in addition to moving images) will in general include at least one object which is of interest and a "background" of lesser interest (and hence of lesser importance).
To analyse the image, e.g. detect the presence/absence or position of a particular object of interest, is often desirable in a variety of applications.
In an image transmission system an improved picture quality might be achieved if data relating to important parts of the scene, i.e. objects of interest, is coded using relatively more bits than data relating to unimportant (i.e. background) parts. For example, in a videophone system a typical image comprises a head and shoulders against a background, and the face area of the head is visually the most important; it is thus desirable to be able to identify the head area from the shoulders and background so as to be able to process the head at a higher refreshment rate than the rest, so that the impression of smooth head motion is conveyed. The ability to locate a head within a head and shoulders scene can thus be used to modify the spatial allocation of video data, enabling a degree of visual importance to be attributed to blocks within the data.
Also, if the position of an object is accurately tracked with time it will be possible to predict its motion, thus allowing "motion compensated" DPCM.
One way of identifying different regions of an image is to utilise the method proposed by Nagao (M. Nagao--"Picture recognition and data structure", Graphic Languages--ed Nake and Rossenfield, 1972). This method has been used in a videophone type system, on an image of a head and shoulders against a background. Some success was achieved in determining the sides of the head when the subject was clean shaven, but very little success was achieved in other cases; so this method is not considered reliable enough for the basis of an area identifying method.
Conventional coders, for instance hybrid discrete cosine transform coders, use no `scene content` information to code the data within the scene, so each part of the scene is operated on as if it has the same visual importance as every other part.
Other image analysis applications are manifold (for example, in automated manufacturing systems).
It is also known to code video images for transmission using Vector Quantisation (VQ). In VQ coding, the image is represented initially by an array of digital data corresponding to the image frame. Blocks of array points ("sub-arrays") are compared with vectors from a codebook, and the best-marching vector selected using a "least squares" difference criterion. A code designating this vector is then transmitted to represent the sub-array. At the receiving end the indicated vector is selected from an identical codebook and displayed.
The underlying principle of the invention, however, is to use VQ as an identification (e.g. object location) method. The extent of the various aspects of the invention are defined in the claims appended hereto.
The different areas of a video image, when vector quantised (VQ), can be operated on differently provided each entry in the VQ codebook has an associated flag indicating which area that entry represents. So in the example of the videophone two different flag entries are requires, one for the head and the other for the remainder of the scene.
An embodiment off the invention will now be described by way of non-limitative example concerned with the identification of a head in a head and shoulders against a background scene, with reference to the accompanying drawings in which:
FIG. 1 is a block diagram illustrating the initial stages of operation of parts of a coder embodying the invention;
FIGS. 2a-g show schematically various stages in the training sequence used to derive the codebook;
FIG. 3 is a block diagram illustrating the operation of a coder embodying the invention;
FIG. 4a shoes schematically a frame to be analysed;
FIG. 4b illustrates the sub-array blocks used in vector quantising FIG. 4a;
FIG. 4c shows the state of flags corresponding to the vector quantised image of FIG. 4b;
FIG. 4d shows schematically the result of analysing the frame of FIG. 4a according to the invention;
FIGS. 5a and 5b show schematically a coder embodying the invention; and
FIG. 5b schematically shows the differential coder of FIG. 5a in more detail.
To enable the invention to operate, it is necessary to have provided a composite codebook which includes vectors flagged as being "head". Preferably others are flagged as being "background". It is possible to derive a "standard" codebook for either an average or a given speaker, but to allow flexibility and greater accuracy of identification, this codebook is derived at the start in an initial "training" sequence. A preferred way of implementing such a sequence will now be described.
To generate "head" and "background" parts of the codebook, it is necessary to unambiguously obtain some "head only" data and "background only" data; a crude initial head detection algorithm is required.
Referring to FIGS. 1 and 2, in order to detect the head, digital data representing several contiguous frames of the head and shoulders image are captured; for instance in a store 1. One of these frames is depicted in FIG. 2a. This data does not need to be extremely accurate, but rather representative.
On the assumption that the prime moving areas within the data sequence are directly associated with the head area, frame differencing 2 is applied to the data representing each adjacent pair of frames. This process typically yields a set of difference data for each adjacent pair representing moving areas together with random noise across the whole image area.
For all picture elements (pels) represented by each set of difference data, each pel above a given threshold value of intensity is set to maximum intensity (255) and each pel below the threshold is set to minimum intensity (0). This `thresholding` 3 removes a large quantity of the random noise and some of the moving areas.
Median filtering 4 is next applied to each set of difference data, which very effectively removes most of the remaining random noise, but erodes only small amounts of the moving areas.
The image represented by each set of data at this stage will rarely provide a clear outline of the head; unless the head to background contrast is very high and the movement of the head between adjacent frames is more than one pel. Often only one side and the top of the head may be depicted as shown in FIG. 2b.
Generally, the moving areas will be clustered in regions around the head area, but some isolated clusters may arise due to motion in other area of the image. A clustering process 5 is used to remove some of the isolated clusters: two orthogonal histograms are generated, one representing the number of `moving` pels in the columns of the image represented by the data and one representing the number of moving pels in the rows of the image represented by the data. The first order moments are calculated and the `centre of gravity` of the image determined, as shown in FIG, 2c. A rectangle is then generated, centred on these co-ordinates, of such dimensions that a given percentage of moving area is included within it, see FIG. 2d. The pels remaining outside this rectangle are set to zero intensity, FIG. 2e. By a suitable choice of rectangle isolated clusters are removed by this process.
Constraints are imposed on the selection of the rectangles in order to reduce the occurrence of incorrect rectangles. Since a very small movement of the head between one frame and the next may produce a very small rectangle, the rate of change of size of the rectangle from one set of data to the next is restricted: either each of the boundary lines of the rectangle are constrained to lie within a small distance of the corresponding boundary in the immediately preceding set of data; or the maximum rate of change of the size of the rectangle is linked to the frame difference energy (e.g. the square of the difference data), so id the difference energy is small the change is kept small, but if the difference energy is large the rate of change may be greater.
The rectangle--rectangles are used because they require very few bits of data to define--is then shrunk if necessary, at 6 in FIG. 1, and as shown in FIG. 2f, to become the smallest rectangle that can be placed around the data to enclose all the remaining non-zero pels. This rectangle is assumed to represent an approximate model of the head.
A border is then created, at 7 in FIG. 1, around the final rectangle, as shown in FIG. 2g. This border defines an exclusion zone form where no data will later be taken. This ensures that when the border is applied to the respective frame of the original image the data inside the border will be exclusively head data and the data outside the border will be exclusively background data.
If five frames of data are initially captured in the store 1, then four adjacent pairs of frames are analysed and four sets of data result. After the four borders have been set 7 the head area data and the background area data are extracted from the first four frames of the original image respectively and the Linde-Buso-Grey algorithm is applied to generate a VQ codebook for each area 8, for example, a 9 bit background codebook and 10 bit head codebook (i.e. codebooks containing respectively 29 and 210 entries). The two codebooks are then combined 9 to form one codebook, each entry of which has an associated flag indicating its origin.
Referring now to FIGS. 3 and 4a-d, after this training sequence is completed, the composite codebook is used to locate the head in successive image frames. The VQ coder operates just as it would in a prior art system using VQ as the transmission coding, but for each block of pels coded 10, the code generated will include a flag (for example, the first digit) indicating whether that block is "head" or "background" so that the head position is known for each frame.
It will of course be appreciated that when the codebook is derived at the coder as indicated above, VQ cannot be used as the transmission code (unless this codebook is made known to the decoder first by transmitting an indication of the vectors).
Since the quantisation process is inherently approximate, it will be appreciated that occasionally blocks from the head part of the image may best match a vector from the "background" part of the codebook, or vice versa. The actual identification of the head will thus usually involve ignoring isolated "head" blocks using erosion and clustering 11, 12 (for example, as described above), or designating the area with the highest concentration of "head" blocks as the actual head.
Another method involves detecting "head" blocks and then examining the error between the block and the "head" vector, and that between the block and the best-matching "background" vector, and if the two scores are similar (i.e. there is ambiguity about whether the block is "head" or "background"), reflagging the block to "background" instead.
If the head blocks are too scattered, it may be that the codebook is insufficient to characterise the head. In this case, a retraining sequence may be employed to regenerate the codebook.
This retraining sequence may either simply be a further sequence of the kind described above, or it may attempt to improve (rather than simply redefine) the codebook. For example, a count may be kept of the number of "incorrect" (i.e. scattered) as opposed to "correct" (i.e. concentrated in the head area) occurences of each vector, and the scatter may thus be reduced by rejecting from the codebook vectors which occur incorrectly too often.
Or, alternatively, the approximate head location derived by locating the greatest concentration of "head" blocks may be used, in the same manner as described above, as an area for generating a new "head" codebook.
These latter approaches, in which VQ coder "learns" from each retraining sequence, are preferred on grounds of accuracy.
FIG. 5a shows a clock diagram of a video coding apparatus (e.g. for a video telephone) embodying the invention. Video signals are fed from an input 20 to a frame store 21 where individual picture element values are recorded in respective store locations so that desired sub-arrays of pels are accessible for further processing. The sub-array sizes may typically be 8×8. In an initial, training, phase of the apparatus a training control unit 22--which may for example be a suitably programmed microprocessor system--carries out the codebook generation method described above, and enters the vectors (and flags) in a VQ codebook store 23. It will be understood that the VQ process involves matching 8×8 sub-array to the nearest one of the stored vectors, viz. a number of 8×8 patterns which are consistently fewer in number than the maximum possible number (264) of such patterns.
In the coding phase of the apparatus, the matching is carried out by VQ control logic 24 which receives successive sub -arrays from the frame store 21 and compares each of these with all the vectors in the codebook store. The simplest form of comparison would be to compute the mean square difference between the two; the vector giving the lowest result being deemed to be the best match. The output from the VQ control logic is the sequence of flags associated with the vectors thus identified.
The actual logic is carried out in this example by an inter frame differential coder 25, in which an inter-frame difference (in subtractor 26) is taken (in conventional manner) between pels from the frame store 21 and a previous frame predictor 27. As is usual in such systems, a quantizer 28 and an output buffer 29 (to match the irregular rate of data generation to a transmission link operating at a constant rate) are shown. A receiver (not shown) used the difference-information to update a reconstructed image in a frame store. The flag output from the VQ control logic 24 is connected (if required, via erode/cluster circuits 30) to the differential coder 25. When the flag indicates that "head" information is being processed, the coder operates normally. If however "background" is indicated, then the generation of difference information is carried out less frequently (e.g. on alternate frames only). This operation is illustrated by a switch 31 which, when the flag indicates "background", breaks the coding loop on alternate frames.
It will be apparent from the foregoing that any visually distinctive object or objects may be accurately detected, recognised or located using methods according to the invention.
While the invention has been described in connection with what is presently considered to be the most practical and preferred embodiment, it is to be understood that the invention is not to limited to the disclosed embodiment, but on the contrary, is intended to cover various modifications and equivalent arrangements included within the spirit and scope of the appended claims.
|Cited Patent||Filing date||Publication date||Applicant||Title|
|US3761613 *||Jun 20, 1972||Sep 25, 1973||Bell Telephone Labor Inc||Dual mode video encoder|
|US3805238 *||May 9, 1972||Apr 16, 1974||Rothfjell R||Method for identifying individuals using selected characteristic body curves|
|US4119946 *||May 20, 1977||Oct 10, 1978||National Research Development Corporation||Comparison apparatus, e.g. for use in character recognition|
|US4295121 *||Nov 19, 1979||Oct 13, 1981||International Business Machines Corporation||Device for optical character reading|
|US4334241 *||Apr 14, 1980||Jun 8, 1982||Hitachi, Ltd.||Pattern position detecting system|
|US4613894 *||Aug 15, 1984||Sep 23, 1986||Thomson-Csf||Method and device for detection of moving points in a television image for digital television systems providing bit-rate compression, with conditional-replenishment|
|US4625329 *||Jan 18, 1985||Nov 25, 1986||Nippondenso Co., Ltd.||Position analyzer for vehicle drivers|
|US4700401 *||Aug 27, 1986||Oct 13, 1987||Dest Corporation||Method and apparatus for character recognition employing a dead-band correlator|
|EP0225729A1 *||Nov 10, 1986||Jun 16, 1987||BRITISH TELECOMMUNICATIONS public limited company||Image encoding and synthesis|
|1||*||IEEE International Conference on Acoustics, Speech and Signal Processing, ICASSP 82, Proceedings, 3 5 May 1982, Paris, FR, vol. 1 or 3, IEEE (U.S.), A. Gersho et al: Image Coding Using Vector Quantization , pp. 428 431.|
|2||IEEE International Conference on Acoustics, Speech and Signal Processing, ICASSP 82, Proceedings, 3-5 May 1982, Paris, FR, vol. 1 or 3, IEEE (U.S.), A. Gersho et al: "Image Coding Using Vector Quantization", pp. 428-431.|
|3||*||IEEE International Conference on Communications, ICC 84, Links for the Future, Science, Systems & Services for Communications, 14 17 May 1984, Amsterdam, NL, vol. 1, C. M. Lin et al.: Motion Compensated Inerframe Color Image Coding , pp. 516 520.|
|4||IEEE International Conference on Communications, ICC '84, Links for the Future, Science, Systems & Services for Communications, 14-17 May 1984, Amsterdam, NL, vol. 1, C. M. Lin et al.: "Motion Compensated Inerframe Color Image Coding", pp. 516-520.|
|5||IEEE International Conference on Communications, Jun. 1982, vol. 1, (Philadelphia, U.S.), F. May et al.: "Picture Coding with Motion Analysis for Low Rate Transmission", pp. 2G, 7.1-2G, 7.5.|
|6||*||IEEE International Conference on Communications, Jun. 1982, vol. 1, (Philadelphia, U.S.), F. May et al.: Picture Coding with Motion Analysis for Low Rate Transmission , pp. 2G, 7.1 2G, 7.5.|
|7||*||Reconnaissance des Formes et Traitement des Images, congres AFCET IRIA, vol. 2, Feb. 1978, (Paris, FR), G. E. Lowitz: Compression des donnees images par reconnaissance des formes et clustering: pp. 699 714.|
|8||Reconnaissance des Formes et Traitement des Images, congres AFCET-IRIA, vol. 2, Feb. 1978, (Paris, FR), G. E. Lowitz: Compression des donnees images par reconnaissance des formes et clustering:--pp. 699-714.|
|9||The Radio & Electronics Engineer, vol. 48, No. 6, Jun. 1978, Institution of Electronics and Radio Engineers, (London, GB) R. C. Nicol: "Interfacing Intraframe d.p.c.m. with Conditional Replenishment Coding for Viewphone Signals", pp. 277-284.|
|10||*||The Radio & Electronics Engineer, vol. 48, No. 6, Jun. 1978, Institution of Electronics and Radio Engineers, (London, GB) R. C. Nicol: Interfacing Intraframe d.p.c.m. with Conditional Replenishment Coding for Viewphone Signals , pp. 277 284.|
|Citing Patent||Filing date||Publication date||Applicant||Title|
|US5187574 *||Aug 21, 1991||Feb 16, 1993||Kanda Tsushin Kogyo Co., Ltd.||Method for automatically adjusting field of view of television monitor system and apparatus for carrying out the same|
|US5218648 *||Dec 17, 1990||Jun 8, 1993||Hughes Aircraft Company||Constellation matching system and method|
|US5243418 *||Nov 27, 1991||Sep 7, 1993||Kabushiki Kaisha Toshiba||Display monitoring system for detecting and tracking an intruder in a monitor area|
|US5386482 *||Jul 16, 1992||Jan 31, 1995||Scan-Optics, Inc.||Address block location method and apparatus|
|US5420638 *||Apr 1, 1993||May 30, 1995||U.S. Philips Corporation||Subassembly for coding images with refresh correction of the data to be coded, and subassembly for decording signals representing these images and previously coded by means of a subassembly of the former kind|
|US5434617 *||Dec 7, 1994||Jul 18, 1995||Bell Communications Research, Inc.||Automatic tracking camera control system|
|US5570434 *||Feb 25, 1994||Oct 29, 1996||U.S. Philips Corporation||Circuit arrangement for recognizing a human face|
|US5596362 *||May 31, 1995||Jan 21, 1997||Lucent Technologies Inc.||Low bit rate audio-visual communication having improved face and lip region detection|
|US5610995 *||Jun 6, 1995||Mar 11, 1997||United Parcel Service Of America, Inc.||Method and apparatus for compressing images containing optical symbols|
|US5623587 *||Jun 12, 1995||Apr 22, 1997||Kideo Productions, Inc.||Method and apparatus for producing an electronic image|
|US5631975 *||Oct 19, 1994||May 20, 1997||Koninkl Philips Electronics Nv||Image segmentation device|
|US5666162 *||Oct 30, 1995||Sep 9, 1997||Canon Kabushiki Kaisha||Moving image signal coding apparatus and moving image signal coding control method|
|US5764803 *||Apr 3, 1996||Jun 9, 1998||Lucent Technologies Inc.||Motion-adaptive modelling of scene content for very low bit rate model-assisted coding of video sequences|
|US5790695 *||May 3, 1996||Aug 4, 1998||Sharp Kabushiki Kaisha||Image coding device for coding image signal to reduce the amount of the information in the image|
|US5802494 *||Aug 5, 1996||Sep 1, 1998||Kabushiki Kaisha Toshiba||Patient monitoring system|
|US5812787 *||Jun 30, 1995||Sep 22, 1998||Intel Corporation||Video coding scheme with foreground/background separation|
|US5815209 *||May 12, 1997||Sep 29, 1998||Matsushita Electric Industrial Co., Ltd.||Encoding method, an encoding apparatus, a decoding method and a decoding apparatus for a moving picture|
|US5835641 *||Jan 11, 1996||Nov 10, 1998||Mitsubishi Denki Kabushiki Kaisha||Image pick-up apparatus for detecting and enlarging registered objects|
|US5841470 *||Jun 29, 1993||Nov 24, 1998||British Telecommunications Public Limited Company||Coding and decoding video signals|
|US5889891 *||Nov 21, 1995||Mar 30, 1999||Regents Of The University Of California||Universal codebook vector quantization with constrained storage|
|US5987154 *||Nov 22, 1996||Nov 16, 1999||Lucent Technologies Inc.||Method and means for detecting people in image sequences|
|US6173069||Mar 31, 1998||Jan 9, 2001||Sharp Laboratories Of America, Inc.||Method for adapting quantization in video coding using face detection and visual eccentricity weighting|
|US6307948 *||Jul 24, 1995||Oct 23, 2001||Semiconductor Energy Laboratory Co., Ltd.||Information processing system for audio and visual transmission system|
|US6307962 *||Nov 7, 1997||Oct 23, 2001||The University Of Rochester||Document data compression system which automatically segments documents and generates compressed smart documents therefrom|
|US6307971||Sep 1, 2000||Oct 23, 2001||Sharewave, Inc.||Method and apparatus for digital data compression|
|US6310973 *||Sep 1, 2000||Oct 30, 2001||Sharewave, Inc.||Method and apparatus for digital data compression|
|US6310974||Sep 1, 2000||Oct 30, 2001||Sharewave, Inc.||Method and apparatus for digital data compression|
|US6310975||Sep 1, 2000||Oct 30, 2001||Sharewave, Inc.||Method and apparatus for digital data compression|
|US6310976||Sep 1, 2000||Oct 30, 2001||Sharewave, Inc.||Method and apparatus for digital data compression|
|US6310977||Sep 1, 2000||Oct 30, 2001||Sharewave, Inc.||Method and apparatus for digital data compression|
|US6310978||Sep 1, 2000||Oct 30, 2001||Sharewave, Inc.||Method and apparatus for digital data compression|
|US6310979||Sep 1, 2000||Oct 30, 2001||Sharewave, Inc.||Method and apparatus for digital data compression|
|US6314207||Sep 1, 2000||Nov 6, 2001||Sharewave, Inc.||Method and apparatus for digital data compression|
|US6404900 *||Jan 14, 1999||Jun 11, 2002||Sharp Laboratories Of America, Inc.||Method for robust human face tracking in presence of multiple persons|
|US6549652||Sep 11, 1998||Apr 15, 2003||Cirrus Logic, Inc.||Method and apparatus for reducing noise during lossy transformation processes|
|US6707484||Aug 25, 1999||Mar 16, 2004||Semiconductor Energy Laboratory Co., Ltd.||Information processing system|
|US6775403 *||Feb 2, 2000||Aug 10, 2004||Minolta Co., Ltd.||Device for and method of processing 3-D shape data|
|US6940545||Feb 28, 2000||Sep 6, 2005||Eastman Kodak Company||Face detecting camera and method|
|US7136506||Mar 3, 2003||Nov 14, 2006||Lockheed Martin Corporation||Correlation based in frame video tracker|
|US7158681||Feb 10, 2003||Jan 2, 2007||Cirrus Logic, Inc.||Feedback scheme for video compression system|
|US7181050 *||Sep 13, 2000||Feb 20, 2007||Sharp Laboratories Of America, Inc.||Method for adapting quantization in video coding using face detection and visual eccentricity weighting|
|US7379559||May 28, 2003||May 27, 2008||Trw Automotive U.S. Llc||Method and apparatus for determining an occupant's head location in an actuatable occupant restraining system|
|US7471832||Feb 24, 2004||Dec 30, 2008||Trw Automotive U.S. Llc||Method and apparatus for arbitrating outputs from multiple pattern recognition classifiers|
|US7583294||May 5, 2005||Sep 1, 2009||Eastman Kodak Company||Face detecting camera and method|
|US7629886||Nov 9, 2005||Dec 8, 2009||Axcess International, Inc.||Method and system for networking radio tags in a radio frequency identification system|
|US7760868||Mar 15, 2004||Jul 20, 2010||Semiconductor Energy Laboratory Co., Ltd||Information processing system|
|US7768546||May 12, 2000||Aug 3, 2010||Axcess International, Inc.||Integrated security system and method|
|US7788008 *||Aug 16, 2007||Aug 31, 2010||Automotive Technologies International, Inc.||Eye monitoring system and method for vehicular occupants|
|US7800503||May 11, 2007||Sep 21, 2010||Axcess International Inc.||Radio frequency identification (RFID) tag antenna design|
|US7841120||Jan 10, 2007||Nov 30, 2010||Wilcox Industries Corp.||Hand grip apparatus for firearm|
|US7859551||Feb 25, 2002||Dec 28, 2010||Bulman Richard L||Object customization and presentation system|
|US8364865||Sep 28, 2010||Jan 29, 2013||Microsoft Corporation||Data simulation using host data storage chain|
|US8514261||Jul 9, 2010||Aug 20, 2013||Semiconductor Energy Laboratory Co., Ltd.||Information processing system|
|US8601184||Dec 26, 2012||Dec 3, 2013||Microsoft Corporation||Updating image in buffer via difference data|
|US8638194||Jul 25, 2008||Jan 28, 2014||Axcess International, Inc.||Multiple radio frequency identification (RFID) tag wireless wide area network (WWAN) protocol|
|US20030051255 *||Feb 25, 2002||Mar 13, 2003||Bulman Richard L.||Object customization and presentation system|
|US20040141081 *||Mar 27, 2002||Jul 22, 2004||Morrison David G.||Video communication|
|US20040156549 *||Feb 10, 2003||Aug 12, 2004||Cirrus Logic, Inc.||Feedback scheme for video compression system|
|US20040175019 *||Mar 3, 2003||Sep 9, 2004||Lockheed Martin Corporation||Correlation based in frame video tracker|
|US20040240706 *||May 28, 2003||Dec 2, 2004||Trw Automotive U.S. Llc||Method and apparatus for determining an occupant' s head location in an actuatable occupant restraining system|
|US20050152582 *||Nov 29, 2004||Jul 14, 2005||Samsung Electronics Co., Ltd.||Multiple person detection apparatus and method|
|US20050175243 *||Feb 5, 2004||Aug 11, 2005||Trw Automotive U.S. Llc||Method and apparatus for classifying image data using classifier grid models|
|US20050185845 *||Feb 24, 2004||Aug 25, 2005||Trw Automotive U.S. Llc||Method and apparatus for arbitrating outputs from multiple pattern recognition classifiers|
|US20050196015 *||Mar 2, 2004||Sep 8, 2005||Trw Automotive U.S. Llc||Method and apparatus for tracking head candidate locations in an actuatable occupant restraining system|
|US20050264658 *||May 5, 2005||Dec 1, 2005||Ray Lawrence A||Face detecting camera and method|
|US20060066444 *||Nov 9, 2005||Mar 30, 2006||Axcess, Inc. A Delaware Corporation||Method and system for networking radio tags in a radio frequency identification system|
|US20070205896 *||Mar 2, 2007||Sep 6, 2007||Axcess International Inc.||System and Method for Determining Location, Directionality, and Velocity of RFID Tags|
|US20070280505 *||Aug 16, 2007||Dec 6, 2007||Automotive Technologies International, Inc.||Eye Monitoring System and Method for Vehicular Occupants|
|US20070285241 *||Mar 20, 2007||Dec 13, 2007||Axcess International Inc.||Multi-Tag Tracking Systems and Methods|
|US20080042850 *||May 11, 2007||Feb 21, 2008||Axcess International Inc.||Radio Frequency Identification (RFID) Tag Antenna Design|
|US20100019887 *||Jul 25, 2008||Jan 28, 2010||Axcess International, Inc.||Multiple Radio Frequency Identification (RFID) Tag Wireless Wide Area Network (WWAN) Protocol|
|US20100271471 *||Jul 9, 2010||Oct 28, 2010||Semiconductor Energy Laboratory Co., Ltd.||Information processing system|
|US20130124551 *||Jul 21, 2011||May 16, 2013||Koninklijke Philips Electronics N.V.||Obtaining keywords for searching|
|EP0718788A3 *||Dec 5, 1995||Jul 17, 1996||Eastman Kodak Co||Title not available|
|EP0718807A3 *||Dec 5, 1995||Jul 10, 1996||Eastman Kodak Co||Title not available|
|WO2000016243A1 *||Sep 8, 1999||Mar 23, 2000||Mate - Media Access Technologies Ltd.||Method of face indexing for efficient browsing and searching ofp eople in video|
|U.S. Classification||382/171, 348/77, 375/E07.209, 382/253, 375/E07.083, 375/240.12, 348/171, 375/240.22, 375/E07.263|
|International Classification||H04N7/36, G06T9/00, H04N7/28, G06T7/00, H04N7/14, H04N7/26, H04N1/41, H04N7/32, G06T7/60, G06T1/00|
|Cooperative Classification||H04N19/20, H04N19/503, H04N19/94, G06K9/00261, G06T9/008|
|European Classification||G06T9/00V, H04N7/28, H04N7/26J4, H04N7/36D, G06K9/00F1V|
|Jul 11, 1995||FPAY||Fee payment|
Year of fee payment: 4
|Jul 14, 1999||FPAY||Fee payment|
Year of fee payment: 8
|Jul 23, 2003||FPAY||Fee payment|
Year of fee payment: 12